keystoneml.nodes

learning

package learning

Visibility
  1. Public
  2. All

Type Members

  1. class ApproximatePCAEstimator extends Estimator[DenseVector[Float], DenseVector[Float]] with Logging

    Approximately estimates a PCA model for dimensionality reduction based on an input dataset.

  2. case class BatchPCATransformer(pcaMat: DenseMatrix[Float]) extends Transformer[DenseMatrix[Float], DenseMatrix[Float]] with Logging with Product with Serializable

    Performs dimensionality reduction on an input dataset where each input item is an NxD array and the projection matrix is a DxK array.

  3. class BlockKernelMatrix[T] extends KernelMatrix

    Column-wise block implementation of a kernel matrix.

  4. class BlockLeastSquaresEstimator extends LabelEstimator[DenseVector[Double], DenseVector[Double], DenseVector[Double]] with WeightedNode with CostModel

    Fits a least squares model using block coordinate descent with provided training features and labels

  5. class BlockLinearMapper extends Transformer[DenseVector[Double], DenseVector[Double]]

    Transformer that applies a linear model to an input.

  6. case class BlockStatistics(popCov: DenseMatrix[Double], popMean: DenseVector[Double], jointMean: DenseMatrix[Double], jointMeanRDD: RDD[DenseVector[Double]]) extends Product with Serializable

  7. class BlockWeightedLeastSquaresEstimator extends LabelEstimator[DenseVector[Double], DenseVector[Double], DenseVector[Double]] with WeightedNode

    Train a weighted block-coordinate descent model using least squares

  8. case class ColumnPCAEstimator(dims: Int, numMachines: Option[Int] = scala.None, cpuWeight: Double = 3.8E-4, memWeight: Double = 0.29, networkWeight: Double = 1.32) extends OptimizableEstimator[DenseMatrix[Float], DenseMatrix[Float]] with Product with Serializable

    Estimates a PCA model for dimensionality reduction based on a sample of a larger input dataset.

  9. trait CostModel extends AnyRef

    A trait that represents a known system performance cost model for a solver.

  10. class DenseLBFGSwithL2[T <: Vector[Double]] extends LabelEstimator[T, DenseVector[Double], DenseVector[Double]] with WeightedNode with CostModel

    Class used to solve an optimization problem using Limited-memory BFGS.

  11. case class DistributedColumnPCAEstimator(dims: Int) extends Estimator[DenseMatrix[Float], DenseMatrix[Float]] with CostModel with Product with Serializable

    Estimates a PCA model for dimensionality reduction based on a sample of a larger input dataset, using a distributed PCA algorithm.

  12. class DistributedPCAEstimator extends Estimator[DenseVector[Float], DenseVector[Float]] with CostModel with Logging

    Estimates a PCA model for dimensionality reduction using a distributedQR.

  13. sealed trait GMMInitializationMethod extends Serializable

  14. class GaussianKernelGenerator extends KernelGenerator[DenseVector[Double]] with Serializable with Logging

    Gaussian (RBF) kernel generator.

  15. class GaussianKernelTransformer extends KernelTransformer[DenseVector[Double]] with Serializable

  16. case class GaussianMixtureModel(means: DenseMatrix[Double], variances: DenseMatrix[Double], weights: DenseVector[Double], weightThreshold: Double = 1.0E-4) extends Transformer[DenseVector[Double], DenseVector[Double]] with Product with Serializable

    A Mixture of Gaussians, usually computed via some clustering process.

  17. case class GaussianMixtureModelEstimator(k: Int, maxIterations: Int = 100, minClusterSize: Int = 40, stopTolerance: Double = 1.0E-4, weightThreshold: Double = 1.0E-4, smallVarianceThreshold: Double = 0.01, absoluteVarianceThreshold: Double = 1.0E-9, initializationMethod: GMMInitializationMethod = KMEANS_PLUS_PLUS_INITIALIZATION, seed: Int = 0) extends Estimator[DenseVector[Double], DenseVector[Double]] with Logging with Product with Serializable

    Fit a Gaussian Mixture model to Data.

  18. trait Gradient[T <: Vector[Double]] extends Serializable

    Computes a gradient given vectors as data, labels, and model weights

  19. case class KMeansModel(means: DenseMatrix[Double]) extends Transformer[DenseVector[Double], DenseVector[Double]] with Product with Serializable

    A KMeans assigning transformer

  20. case class KMeansPlusPlusEstimator(numMeans: Int, maxIterations: Int, stopTolerance: Double = 0.001, seed: Int = 0) extends Estimator[DenseVector[Double], DenseVector[Double]] with Logging with Product with Serializable

    Trains a k-means++ transformer

  21. class KernelBlockLinearMapper[T] extends Transformer[T, DenseVector[Double]]

    Transformer that applies a kernel model to an input.

  22. trait KernelGenerator[T] extends AnyRef

    Base trait for functions that compute a kernel on a given dataset.

  23. trait KernelMatrix extends AnyRef

    Defines a wrapper to access elements of a symmetric distributed matrix that is generated using a kernel function.

  24. class KernelRidgeRegression[T] extends LabelEstimator[T, DenseVector[Double], DenseVector[Double]]

    Solves a kernel ridge regression problem of the form (K(x, x) + \lambda * I) * W = Y using Gauss-Seidel based Block Coordinate Descent.

  25. trait KernelTransformer[T] extends AnyRef

    Base trait for functions that transform the given input data with respect to a pre-specified kernel function.

  26. class LeastSquaresDenseGradient extends DenseGradient

    Computes a Least Squares loss gradient given DenseVectors as data

  27. class LeastSquaresEstimator[T <: Vector[Double]] extends OptimizableLabelEstimator[T, DenseVector[Double], DenseVector[Double]] with WeightedNode with Logging

    A least squares solver that is optimized to use a fast algorithm, based on characteristics of the workload and cost models.

  28. class LeastSquaresSparseGradient extends SparseGradient

    Computes a Least Squares loss gradient given SparseVectors as data

  29. class LinearDiscriminantAnalysis extends LabelEstimator[DenseVector[Double], DenseVector[Double], Int]

    An Estimator that fits Linear Discriminant Analysis (currently not calculated in a distributed fashion), and returns a transformer that projects into the new space

  30. class LinearMapEstimator extends LabelEstimator[DenseVector[Double], DenseVector[Double], DenseVector[Double]] with CostModel

    Linear Map Estimator.

  31. case class LinearMapper[T <: Vector[Double]](x: DenseMatrix[Double], bOpt: Option[DenseVector[Double]] = scala.None, featureScaler: Option[StandardScalerModel] = scala.None) extends Transformer[T, DenseVector[Double]] with Product with Serializable

    Computes A * x + b i.

  32. case class LocalColumnPCAEstimator(dims: Int) extends Estimator[DenseMatrix[Float], DenseMatrix[Float]] with CostModel with Product with Serializable

    Estimates a PCA model for dimensionality reduction based on a sample of a larger input dataset.

  33. class LocalLeastSquaresEstimator extends LabelEstimator[DenseVector[Double], DenseVector[Double], DenseVector[Double]]

    Learns a linear model (OLS) based on training features and training labels.

  34. case class LogisticRegressionEstimator[T <: Vector[Double]](numClasses: Int, regParam: Double = 0, numIters: Int = 100, convergenceTol: Double = 1.0E-4, numFeatures: Int = -1)(implicit evidence$1: ClassTag[T]) extends LabelEstimator[T, Double, Int] with Product with Serializable

    A LabelEstimator which learns a Logistic Regression model from training data.

  35. case class LogisticRegressionModel[T <: Vector[Double]](model: org.apache.spark.mllib.classification.LogisticRegressionModel) extends Transformer[T, Double] with Product with Serializable

    A Logistic Regression model that transforms feature vectors to vectors containing the logistic regression output of the different classes

  36. case class NaiveBayesEstimator[T <: Vector[Double]](numClasses: Int, lambda: Double = 1.0)(implicit evidence$1: ClassTag[T]) extends LabelEstimator[T, DenseVector[Double], Int] with Product with Serializable

    A LabelEstimator which learns a multinomial naive bayes model from training data.

  37. class NaiveBayesModel[T <: Vector[Double]] extends Transformer[T, DenseVector[Double]]

    A Multinomial Naive Bayes model that transforms feature vectors to vectors containing the log posterior probabilities of the different classes

  38. class PCAEstimator extends Estimator[DenseVector[Float], DenseVector[Float]] with CostModel with Logging

    Estimates a PCA model for dimensionality reduction based on a sample of a larger input dataset.

  39. class PCATransformer extends Transformer[DenseVector[Float], DenseVector[Float]]

    Performs dimensionality reduction on an input dataset.

  40. class PerClassWeightedLeastSquaresEstimator extends LabelEstimator[DenseVector[Double], DenseVector[Double], DenseVector[Double]]

    Train a weighted block-coordinate descent model using least squares

  41. class SparseLBFGSwithL2 extends LabelEstimator[SparseVector[Double], DenseVector[Double], DenseVector[Double]] with WeightedNode with CostModel

    Class used to solve an optimization problem using Limited-memory BFGS.

  42. case class SparseLinearMapper(x: DenseMatrix[Double], bOpt: Option[DenseVector[Double]] = scala.None) extends Transformer[SparseVector[Double], DenseVector[Double]] with Product with Serializable

    Computes A * x + b i.

  43. class ZCAWhitener extends Transformer[DenseMatrix[Double], DenseMatrix[Double]]

  44. class ZCAWhitenerEstimator extends Estimator[DenseMatrix[Double], DenseMatrix[Double]]

    Computes a ZCA Whitener, which is intended to rotate an input dataset to identity covariance.

Value Members

  1. object ApproximatePCAEstimator extends Serializable

  2. object BlockLeastSquaresEstimator extends Serializable

  3. object BlockWeightedLeastSquaresEstimator extends Logging with Serializable

  4. object GaussianMixtureModel extends Serializable

  5. object Gradient extends Serializable

  6. object KMEANS_PLUS_PLUS_INITIALIZATION extends GMMInitializationMethod with Product with Serializable

  7. object KernelRidgeRegression extends Logging with Serializable

  8. object LBFGSwithL2 extends Logging

  9. object LinearMapEstimator extends Serializable

    Companion object to LinearMapEstimator that allows for construction without new.

  10. object LocalLeastSquaresEstimator extends Serializable

  11. object PCAEstimator extends Serializable

  12. object PerClassWeightedLeastSquaresEstimator extends Logging with Serializable

  13. object RANDOM_INITIALIZATION extends GMMInitializationMethod with Product with Serializable

    Randomly distributes the initial means within the minimum and maximum values seen in the training data, using a uniform distribution.

  14. package external

  15. package internal

Ungrouped